hdf5: add chunk option for row size
Requested by @sole : Hello,
If I take a look at the code, it seems it was not implemented:
To ask for supporting any user desired chunk size might be too cumbersome and rarely used. Nevertheless, at the very least chunking based on the frame rows should be supported. Being able to specify a chunk size (1, n, n_columns) where n is smaller than n_rows would be greatly beneficial for tomography.
We are typically loosing a factor of 10 in reading speed that should be recoverable with a chunking around (1, n_rows/10, n_columns) or even smaller.
Will chunking by n rows be possible in LIMA or we are face to another example of why it should be possible to decouple taking data from a detector and writing a file?